This forum in general have slow respond.
I put in the suggestion/problems box and tagged you there too. Let us see. Thanks for explaining.
ITU method is based upon TCP/IP based subscription, In Indonesia however every SIM card that is newly registered / activated came included with TCP/IP subscription aka the internet. Whether the internet subscription is practically usable or not is another story, what matter is it get count toward the total intenet subscriber even when the subscription plan is practically useless. Also keep in mind that once a SiM card is activated it will be persistent in the statistic. For example, when I dump a SIM card that originally has an internet subscription plan that I no longer use, the SIM card ID (account) will be recycled into a new physical SIM card to be offered to a new customer, however during that recycle period the original subscription plan will persist in the statistic even when the new SIM card sitting on the store shelf waiting for a new customer
48% internet penetration of Indonesia population in 2019 is literally Impossible. My best guess is around 10% of total population. And In less than a decade the figures of Indonesia internet subscriber will be larger than the actual total population
Well there is lot of user survey and household survey too (in ITU), you can read if interested:
Larger site for it: https://www.itu.int/en/ITU-D/Statistics/Pages/definitions/default.aspx
Why I say what I say is ITU definition and surveying is at least fairly standard, so every country more or less gets "rigged" the same amount by say people saying "I use internet" but in somewhat frivolous or rare manner (say because they are merely in household or know household that has internet) in reality compared to ones that are "really" using it daily etc. The kind of thing that madokafc is hinting at in her post.
I say all this (why I hold ITU to be credible) because I remember BTRC for Bangladesh put their internet reach data as ridiculously high (using IP = person correlation precisely as you state it, which is bad assumption), but under ITU definition (by ground person survey) it comes to 15% which make lot more sense in standardised sense w.r.t their income level.
Anyway we have made our points, I understand where you are coming from. There is problem with lot of statistics in general for sure....there is overall bad ones and less-bad ones...and we can discuss why the less bad ones are less bad lol.